How ranking works

We collect Grok Bot prompts and setups shared on X, rank them by how often people bookmark them, and set aside bait, promo campaigns and sales pitches.

Popularity, not a test

The save score shows what people on X found worth keeping. We don't run the prompts, so a high score isn't proof a prompt works. Read it before you use it.

The save score

Save score = 100 × ( 0.5·save rate + 0.3·volume + 0.2·save ratio )

50%

Save rate

(bookmarks + 50) ÷ (views + 5,000)

3% or more = full marks. Smoothed so tiny posts can't fluke it.

30%

Volume

log₁₀(1 + bookmarks) ÷ log₁₀(5,001)

5,000+ bookmarks = full marks. Log scale, so 100→1K counts like 1K→10K.

20%

Save ratio

bookmarks ÷ (bookmarks + likes)

Saved more than liked usually means a tool, not applause.

Flags never change the score. They decide whether a post is shown by default.

Two collections

Prompts & templatesPosts with a real prompt you can read: the full text, or a substantial part of it. If a post links to a free, public page, we copy the prompt verbatim and link the source.
Guides & resourcesReading lists, talks, repos and posts with no copy-paste prompt.

Each collection is ranked on its own, so #1 is always the highest score in that list. This week's picks come only from Prompts & templates.

What gets benched

Benched posts are hidden by default. Turn on "Include bait, promo and seller posts" under Filters to see them, with the reason.

🎣 Bait"Comment X to get it," follow-gating, "steal this before…"
📢 Promo campaignNear-identical text posted by 3+ accounts.
$ Sells somethingPaid product, agency pitch, "DM me" or a private group funnel, in the post or the page it links to.
? Attribution unverifiedCredits a famous name or photo that didn't write it.
Link onlyAlmost no text, just a link.
🔥 Hype wordingIncome claims or "95% of people…". Shown in details, not benched.

Prompt availability

Full promptThe whole prompt is in the post, its image, or a free linked page. Copy it in one click.
PartialOnly part is public. Treat it as a starting point.

Setup time

Easy · <1h0–1 points
Medium · 1–3h2–3 points
Hard · 3h+4+ points

+1 per connector beyond the first · +1 new account or API key · +2 code, GitHub or self-hosting · +1 paid tool · +1 for 4+ bots, +1 more for 10+. It's an estimate from the post, not a timed run.

Rules run automatically on new posts, and the top of each list is checked by hand. Spot a mistake? Tell us on X at @thebotbench.